[Samba] Samba CTDB clustering panic

[Martin Schwenke]

Hi João,

On Tue, 6 Aug 2019 14:54:32 +0100, João Baúto via samba
<samba at lists.samba.org> wrote:

> I'm setting up a clustered environment on CentOS 7.6.1810 with
> Samba 4.8.3-6.el7_6 and the corresponding ctdb package. I'm using FreeIPA
> 4.6.4 as my auth backend.
> 
> Currently I'm running into this bug
> https://bugzilla.samba.org/show_bug.cgi?id=13465 however I would assume
> that it is fixed in the latest release (dated 29July 2019). It crashed once
> CTDB runs testparm to check the configuration. Setting clustering to no
> allows access to the shares via public address.
> 
> The panic message is,
> 
> 50.samba: /etc/ctdb/events.d/50.samba: line 79: 15459 Aborted
>   timeout "$_timeout" testparm -v -s > "$_out" 2> "$_err"
> 50.samba: ERROR: smb.conf cache create failed - testparm failed with:
> 50.samba: Load smb config files from /etc/samba/smb.conf
> 50.samba: rlimit_max: increasing rlimit_max (1024) to minimum Windows limit
> (16384)
> 50.samba: Processing section "[swp]"
> 50.samba: smb_panic_default: PANIC (pid 15460): messaging not initialized
> 50.samba: BACKTRACE: 17 stack frames:
> 50.samba:  #0 /lib64/libsamba-util.so.0(log_stack_trace+0x1a)
> [0x7f599afd9aaa]
> 50.samba:  #1 /lib64/libsamba-util.so.0(smb_panic+0x52) [0x7f599afd9bb2]
> 50.samba:  #2 /usr/lib64/samba/libsamba-cluster-support-samba4.so(+0x77a0)
> [0x7f5997dab7a0]
> 50.samba:  #3 /lib64/libsmbconf.so.0(db_open+0x253) [0x7f599b26db93]
> 50.samba:  #4 /lib64/libsmbconf.so.0(regdb_init+0x86) [0x7f599b255566]
> 50.samba:  #5 /lib64/libsmbconf.so.0(registry_init_common+0x6)
> [0x7f599b256e46]
> 50.samba:  #6 /lib64/libsmbconf.so.0(registry_init_smbconf+0x26)
> [0x7f599b285d76]
> 50.samba:  #7 /lib64/libsmbconf.so.0(+0x56c16) [0x7f599b284c16]
> 50.samba:  #8 /lib64/libsmbconf.so.0(smbconf_init_internal+0x49)
> [0x7f599b2877a9]
> 50.samba:  #9 /lib64/libsmbconf.so.0(smbconf_init+0x120) [0x7f599b284000]
> 50.samba:  #10 /lib64/libsmbconf.so.0(+0x29146) [0x7f599b257146]
> 50.samba:  #11 /lib64/libsmbconf.so.0(process_registry_shares+0x2f)
> [0x7f599b260dbf]
> 50.samba:  #12 /lib64/libsmbconf.so.0(+0x34a2a) [0x7f599b262a2a]
> 50.samba:  #13 /lib64/libsmbconf.so.0(lp_load_with_registry_shares+0x22)
> [0x7f599b262c82]
> 50.samba:  #14 testparm(main+0x3a2) [0x55a45fc15b32]
> 50.samba:  #15 /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f599a1d0495]
> 50.samba:  #16 testparm(+0x3f1c) [0x55a45fc16f1c]
> 50.samba: timeout: the monitored command dumped core
> monitor event failed
> 
> *smb.conf*
> [global]
> netbios name = EXAMPLE
> realm = EXAMPLE.COM
> workgroup = EXAMPLE
> domain master = Yes
> ldap group suffix = cn=groups,cn=accounts
> ldap machine suffix = cn=computers,cn=accounts
> ldap ssl = off
> ldap suffix = dc=example,dc=com
> ldap user suffix = cn=users,cn=accounts
> log file = /var/log/samba/log
> dedicated keytab file = FILE:/etc/samba/samba.keytab
> kerberos method = dedicated keytab
> passdb backend = ipasam:ldap://ipa.example.com
> security = USER
> idmap config * : backend = tdb
> ldap admin dn = cn=Directory Manager
> writable = yes
> browseable = yes
> read only = no
> guest ok = no
> clustering = yes
> 
> [test]
> create mode = 0640
> directory mode = 0750
> path = /gluster/data
> valid users = john.doe
> 
> Anyone has an idea why this is happening?

Assuming the CTDB event script is running as root (which is the
default... and all that is really possible/recommended), the fix for
13465 is in Samba v4-8-stable:

  $ git describe 8d0fbe3ebb134bbf6cbdfdf86e0b0279ae8d1c73
  samba-4.8.6-23-g8d0fbe3ebb1

This means that it appeared in samba-4.8.7.  The fix may not have been
backported to the CentOS 7 version of Samba (4.8.3-6.el7_6).  In this
case you would need to report a bug to CentOS.

Please also see:

  https://bugzilla.samba.org/show_bug.cgi?id=13925

Good luck!

peace & happiness,
martin